Optimized stacked long short-term memory with hyperbolic secant activation function for Alzheimer’s disease classification

Alzheimer’s Disease (AD) is a progressive disorder that results in reflective failure in human memory and cognition. The existing methods have a high learning rate which affects the model stability and generates poor generalizability thereby reducing classifier accuracy. Therefore, the Stacked Long...

Full description

Saved in:
Bibliographic Details
Published in:Biomedical signal processing and control Vol. 108; p. 107980
Main Authors: Maaram, Krishna Kishore, Chandre, Shanker
Format: Journal Article
Language:English
Published: Elsevier Ltd 01.10.2025
Subjects:
Alzheimer’s disease
Crisscross Strategy
Hyperbolic Secant
Parrot Optimization Algorithm
Stacked Long Short-Term Memory
Parrot Optimization Algorithm
Stacked Long Short-Term Memory
Alzheimer’s disease
Crisscross Strategy
Hyperbolic Secant
ISSN:1746-8094
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Alzheimer’s Disease (AD) is a progressive disorder that results in reflective failure in human memory and cognition. The existing methods have a high learning rate which affects the model stability and generates poor generalizability thereby reducing classifier accuracy. Therefore, the Stacked Long Short-Term Memory (SLSTM) with hyperbolic secant (Sech) activation function is proposed in this research for classifying AD diseases. To enhance the classifier performance, the hyperparameter tuning is applied by using the Crisscross Strategy based Parrot Optimization Algorithm (CSPOA). The sech function has a moderate slope which potentially leads to reduced vanishing gradient issues. The hidden layer of SLSTM creates a model for deeper and more accurate learning of description. The model parameters are spread in the entire space without enhancing memory capacity thereby enabling convergence and nonlinear operations of data. The flipping is applied for AD, Cognitively Normal (CN) and Mild Cognitive Impairment (MCI) classes of Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET) data to augment the images and min–max normalization is applied for augmented AD, CN and MCI classes. The optimized SLSTM with Sech based classifier attained 99.18% accuracy for the class of AD vs CN, 99.52% accuracy for MCI vs CN, 99.17% accuracy on MCI vs AD and 98.25% accuracy for AD vs CN vs MCI which is better than Multimodal image feature fusion.
ISSN:1746-8094
DOI:10.1016/j.bspc.2025.107980